Description
Camera-based observation of a venue for safety, security, or analytics purposes. Video surveillance is the dominant baseline against which wireless human sensing is positioned in the literature: cameras give high-fidelity data but raise privacy issues that wireless sensing aspires to avoid. For this thesis, video surveillance is treated as the comparison category — the benchmark accuracy ceiling for crowd-counting / crowd-monitoring and the privacy regime that BLE-calibrated CSI is supposed to be cleaner than.
Why it's relevant here
- Many CSI / BLE counting papers benchmark against vision systems trained on the same scenes.
- Most "intelligent" / "AI-based" crowd-analysis surveys are vision-anchored — the dominant tradition that wireless sensing must engage with.
- Privacy regulation (GDPR, surveillance acts) motivates wireless alternatives.
- Vision baselines bound what is recoverable about a scene; wireless techniques inherit aspirational accuracy targets from them.
Why it's hard (from a research-positioning angle)
- Camera infrastructure is widely deployed; arguing for a wireless alternative requires demonstrating a privacy or coverage benefit.
- Vision benchmarks dominate dataset culture; wireless-only datasets are still rare and non-standardized.
- Hybrid camera + wireless systems are common in industry but rarely studied as such in academia.
Source Papers
- sreenu2019_6f76 ↗ — intelligent video surveillance: review through deep learning techniques.
- bendalibraham2021_476e ↗ — recent trends in crowd analysis (vision-dominant).
- sindagi2018_e579 ↗ — CNN-based crowd counting and density estimation (vision baseline).
- davies1995_b3cd ↗ — early crowd monitoring using image processing.